Better hack for supporting user using OCRmyPDF inside WSL (#2)

Author: antoinelibert

Dockerfile

@@ -40,4 +40,7 @@ COPY api/ /app/api
 
 EXPOSE 8000
 
+ENV WORKDIR /workdir
+VOLUME /workdir
+
 CMD ["uvicorn", "api.main:app", "--host", "0.0.0.0", "--port", "8000"]

api/main.py

@@ -66,7 +66,7 @@ def clean_docs():
 
 def do_ocr(_doc: Document):
     pool_ocr.acquire()
-    _doc.ocr()
+    _doc.ocr(config.enable_wsl_compat)
     pool_ocr.release()
 
 

api/models.py

@@ -8,6 +8,7 @@
 from pydantic import BaseModel
 
 from api.settings import config
+from api.tools import special_win_wslpath
 
 
 class Lang(str, Enum):
@@ -43,20 +44,33 @@ class Document(BaseModel):
     expire: datetime
     finished: Optional[datetime] = None
 
-    def ocr(self):
+    def ocr(self, wsl: bool = False):
         self.status = "processing"
         self.processing = datetime.now()
         self.save_state()
+
+        # Hack for user using OCRmyPDF inside WSL (Windows)
+        output_txt_path = (
+            special_win_wslpath(self.output_txt)
+            if wsl
+            else str(self.output_txt.absolute())
+        )
+        input_path = (
+            special_win_wslpath(self.input) if wsl else str(self.input.absolute())
+        )
+        output_path = (
+            special_win_wslpath(self.output) if wsl else str(self.output.absolute())
+        )
         try:
             output = subprocess.check_output(
                 " ".join(
                     [
                         config.base_command_ocr,
                         config.base_command_option,
                         f"-l {'+'.join([l.value for l in self.lang])}",
-                        f"--sidecar {self.output_txt.resolve().relative_to(config.basedir).as_posix()}",
-                        self.input.resolve().relative_to(config.basedir).as_posix(),
-                        self.output.resolve().relative_to(config.basedir).as_posix(),
+                        f"--sidecar {output_txt_path}",
+                        input_path,
+                        output_path,
                     ]
                 ),
                 stderr=subprocess.STDOUT,

api/settings.py

@@ -12,6 +12,7 @@ class Settings(BaseSettings):
     base_command_option: str = "--output-type pdf --fast-web-view 0 --optimize 0"
     max_ocr_process: int = 1
     document_expire_hour: int = 1
+    enable_wsl_compat: bool = False
 
 
 config = Settings()

api/tools.py

@@ -10,3 +10,10 @@ def save_upload_file(upload_file: UploadFile, destination: Path) -> None:
             shutil.copyfileobj(upload_file.file, buffer)
     finally:
         upload_file.file.close()
+
+
+def special_win_wslpath(path: Path) -> str:
+    """
+    This is a special function returning a compatible path for user of OCRmyPDF inside WSL
+    """
+    return f"`wslpath -a '{str(path)}'`"

tests/test_models.py

@@ -48,13 +48,9 @@ def test_ocr(self, monkeypatch, tmp_path, document_model, subprocess_check_outpu
                     api.settings.config.base_command_ocr,
                     api.settings.config.base_command_option,
                     f"-l {'+'.join([l.value for l in document.lang])}",
-                    f"--sidecar {document.output_txt.resolve().relative_to(api.settings.config.basedir).as_posix()}",
-                    document.input.resolve()
-                    .relative_to(api.settings.config.basedir)
-                    .as_posix(),
-                    document.output.resolve()
-                    .relative_to(api.settings.config.basedir)
-                    .as_posix(),
+                    f"--sidecar {str(document.output_txt.absolute())}",
+                    str(document.input.absolute()),
+                    str(document.output.absolute()),
                 ]
             ),
             stderr=subprocess.STDOUT,